CN116029957A - Insulator image pollution identification method based on Markov chain Monte Carlo - Google Patents
Insulator image pollution identification method based on Markov chain Monte Carlo Download PDFInfo
- Publication number
- CN116029957A CN116029957A CN202111029250.4A CN202111029250A CN116029957A CN 116029957 A CN116029957 A CN 116029957A CN 202111029250 A CN202111029250 A CN 202111029250A CN 116029957 A CN116029957 A CN 116029957A
- Authority
- CN
- China
- Prior art keywords
- image
- pollution
- insulator
- insulator image
- sample
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Insulators (AREA)
- Image Analysis (AREA)
- Testing Relating To Insulation (AREA)
Abstract
The invention discloses an insulator image pollution identification method based on Markov chain Monte Carlo, which comprises the following steps: s1, establishing an insulator image library; s2, dividing an insulator image in an insulator image library of the distribution line into a training set and a verification set; s3, establishing an insulator image pollution identification sample set; s4, establishing an insulator image pollution identification feature set; s5, constructing a Bayes convolutional neural network classifier, and initializing Bayes convolutional neural network parameters by using an Xavier method; s6, training by adopting a Markov chain Monte Carlo method to obtain an insulator image pollution recognition model; s7, detecting and identifying pollution types of the sub-images of the insulators. According to the invention, the Markov chain Monte Carlo method is adopted to obtain the pollution recognition model of the insulator image, so that the efficiency of the insulator image detection and the pollution recognition is improved, and the effectiveness and the accuracy of the pollution recognition of the insulator image are ensured.
Description
Technical Field
The invention belongs to the technical field of distribution line equipment running state maintenance and computer vision, and particularly relates to an insulator image pollution identification method based on Markov chain Monte Carlo.
Background
Transmission lines, which carry large currents from a power plant to a remote user, must have two basic conditions, one of which is to provide mechanical support for the conductors carrying the current; secondly, the insulator has two basic functions of preventing the current from forming a channel to be grounded, so the insulator plays an important role in a power transmission line.
The pollution on the surface of the insulator easily causes the reduction of the external insulation performance of equipment in the power industry, thereby causing a plurality of safety faults. The traditional manual inspection mode has the problems of efficiency, safety and the like, is easily affected by local terrain and climate, and has higher inspection difficulty.
In addition, the traditional insulator image pollution recognition method is generally based on color shape characteristics, is easily influenced by uncertain factors such as illumination, weather, complex image background and the like, and has low recognition accuracy. The neural network based on the Bayesian probability model adds probability distribution on model parameters and model output to consider the influence of uncertainty factors, but in the process of searching the optimal approximate distribution by a common variation inference method, result deviation is easy to cause, and accuracy is reduced, so that an accurate and efficient insulator image pollution identification method needs to be researched.
Disclosure of Invention
In order to solve the technical problems, the invention provides the method for identifying the filth of the insulator image based on the Markov chain Monte Carlo, which adopts the Markov chain Monte Carlo method to obtain the filth identification model of the insulator image, so that the efficiency of the insulator image detection and filth identification is improved, and the effectiveness and the accuracy of the insulator image filth identification are ensured.
The technical scheme provided by the invention is as follows:
the method for identifying the insulator image pollution based on the Markov chain Monte Carlo specifically comprises the following steps:
s1, establishing an insulator image library: collecting and storing the data of the insulation sub-images of the distribution line by using unmanned aerial vehicle inspection;
s2, dividing an insulator image in an insulator image library of the distribution line into a training set and a verification set;
s3, establishing an insulator image pollution identification sample set: manually screening the training set in the step S2, and classifying whether the insulator image has a pollution mark or not to obtain an insulator image pollution identification sample set, namely an insulator image pollution sample set and an insulator image pollution-free sample set;
s4, establishing an insulator image pollution identification feature set: each insulator image in the insulator image pollution identification sample set is subjected to segmentation pretreatment, feature extraction and selection, and the features are combined into a feature vector Y to obtain an insulator image pollution identification feature set, and the insulator images are divided into a polluted insulator image and a non-polluted insulator image according to the feature vector Y;
s5, constructing a Bayes convolutional neural network classifier, and initializing Bayes convolutional neural network parameters by using an Xavier method: a. determining a distribution model of an image characteristic to obtain prior probability and class conditional probability of a sample image; b. calculating the corresponding posterior probability by using a Bayes formula; c. obtaining an image retrieval result set meeting the requirements by utilizing a decision function J (X);
s6, training by adopting a Markov chain Monte Carlo method to obtain an insulator image pollution recognition model: training model parameters by adopting a numerical sampling Markov chain Monte Carlo method to obtain probability distribution of neural network parameters, and optimizing the model parameters to obtain an insulator image pollution recognition model;
s7, detecting and identifying pollution types of the sub-images of the insulators: and (3) detecting the verification set in the step (2) by using the insulator image pollution identification model in the step (S6), respectively calculating the probability of pollution and no pollution of the insulator image in the verification set, and selecting the probability with a large value to judge the probability as the pollution type of the insulator image.
Further, the specific operation in the step S5The method comprises the following steps: let the adopted insulator image pollution sample set be R = { R 1 ,R 2 ,...,R k The sample set of the insulator image without pollution is N = { N = 1 ,N 2 ,...,N k Probability can be estimated using statistical methods:
the probability of the insulator image being dirty is that the image X is:
P(X|R)≈P(X|R*) (1)
the probability of the image X being an insulator image without contamination is as follows:
P(X|N)≈P(X|N*) (2)
according to the Bayesian formula: j (X) = -log [ P (r|x) ]+log [ P (n|x) ], wherein P (r|x) and P (n|x) can be obtained by posterior probability learning of bayesian formulas, i.e., P (r|x) and P (n|x) use a markov chain monte carlo method based on numerical sampling to obtain an approximate solution of the posterior distribution.
The smaller the value of J (X) in the Bayesian formula is, the more the required retrieval image X meets the requirements.
Since P (X) is typically a constant, J (X) can be transformed into:
J(X)'=-log[P(X|R)P(R)]+log]P(X|N)P(N)] (3)
setting a threshold value alpha, and when the obtained J (X)' < alpha, obtaining the detected image which is the insulator image with the dirt.
Further, the specific steps of initializing the bayesian convolutional neural network parameters by using the Xavier method in the step S5 are as follows:
a. generating a normal distribution random matrix with the mean value of 0 and the standard deviation of 0.01;
b. judging whether the generated matrix meets the requirement of the formula (1) or not by using the formula (2), if not, continuously judging the regenerated matrix until the distribution of the weights meets the consistent distribution of the formula (1), namely
W in the formula (4) is weight distribution, U is uniform distribution, n j Is of layer jNumber of neurons.
wherein x is as described above m For point set x n (n=1, 2,.+ -.) j is a j-th layer neural network, h=0, ±1, ±2...
Further, the specific algorithm of the markov chain monte carlo method in the step S6 is as follows: consider the sampling process as a markov chain: x is x 1 ,x 2 ,...,x t-1 ,x t ,x t+1 ,. x in the formula t Representing the sample of the t-th sample, the sample of the t+1th sample depends on the sample of the t-th sample and the state transition distribution q (x|x t ). If the plateau distribution of the Markov chain is p (x), then the samples at the plateau of the state follow the p (x) distribution.
Specifically, according to the state transition distribution q (x|x t ) Extracting a sample x 'and estimating a probability A (x', x t ) To take x' as sample x of the (t+1) th sample t+1 。
Due to each time q (x|x t ) Randomly extracting a sample and taking the sample as A (x', x t ) The probability of accepting, the modified markov chain state transition probability is:
q'(x'|x t )=q(x'|x t )A(x',x t ) (7)
the fine and smooth conditions according to the Markov chain are:
the modified markov chain can reach a plateau and a plateau distribution is p (x).
The beneficial effects of the invention are as follows:
1) The unmanned aerial vehicle is adopted to patrol the insulator image, so that the efficiency and the definition of the acquired image are improved;
2) The technical problems that the model parameters and model output are affected by uncertain factors such as illumination, weather, complex picture background and the like easily are solved by adding probability distribution to the model parameters and the model output through a convolutional Bayesian neural network;
3) The adopted Markov chain Monte Carlo method randomly extracts samples from the real distribution to generate probability distribution of parameters by sampling, and the probability distribution of each parameter in the neural network is regarded as a hidden variable, so that compared with the variation method in the prior art, the method is simpler and faster, omits complicated mathematical deduction and calculation processes, and is easier to realize in a computer program;
4) Compared with a variational Bayesian method and a support vector machine method, the adopted Markov chain Monte Carlo method is more effective and has higher accuracy, and can rapidly identify and classify the pollution of the insulator image.
Drawings
FIG. 1 is a flow diagram of a Markov chain Monte Carlo based method for identifying a filth of an insulator image of the present invention;
FIG. 2 is a schematic flow chart of the algorithm of the Markov chain Monte Carlo method in the present invention;
FIG. 3 is a schematic diagram of a 10kV Jiang Bu I line artificially marked pole tower insulation sub-image pollution sample;
FIG. 4 is a schematic diagram of the effect of identifying the pollution of the sub-image of the insulator in the present invention;
fig. 5 is a schematic diagram of classification accuracy curves of the variational bayes method and the markov chain monte carlo method in the training process.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.
It should be noted that, in the description of the present invention, it should be noted that, as the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. are shown based on the orientation or the positional relationship shown in the drawings, only for convenience of describing the present invention and simplifying the description, and do not indicate or imply that the apparatus or elements to be referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as limiting the present invention.
Example 1
As shown in fig. 1 to 4, wherein the horizontal axis and the vertical axis in fig. 3 and 4 each represent the number of pixels, the method for identifying the filth of the insulator image based on the markov chain monte carlo comprises the following steps:
s1, establishing an insulator image library: collecting and storing the data of the insulation sub-images of the distribution line by using unmanned aerial vehicle inspection;
s2, dividing an insulator image in an insulator image library of the distribution line into a training set and a verification set;
in this embodiment, 80% of the insulator images in the insulator image library are used as the training set, and 20% of the insulator images are used as the verification set.
S3, establishing an insulator image pollution identification sample set: manually screening the training set in the step S2, and classifying whether the insulator image has a pollution mark or not to obtain an insulator image pollution identification sample set, namely an insulator image pollution sample set and an insulator image pollution-free sample set;
s4, establishing an insulator image pollution identification feature set: each insulator image in the insulator image pollution identification sample set is subjected to segmentation pretreatment, feature extraction and selection, and the features are combined into a feature vector Y to obtain an insulator image pollution identification feature set, and the insulator images are divided into a polluted insulator image and a non-polluted insulator image according to the feature vector Y;
in this embodiment, the specific algorithm in step S4 is as follows: let the adopted insulator image pollution sample set be R = { R 1 ,R 2 ,...,R k The sample set of the insulator image without pollution is N = { N = 1 ,N 2 ,...,N k Probability, i.e., using statistical methods, can be estimated
The probability of the insulator image being dirty is that the image X is:
P(X|R)≈P(X|R*) (1)
the probability of the image X being an insulator image without contamination is as follows:
P(X|N)≈P(X|N*) (2)
according to the Bayesian formula: j (X) = -log [ P (r|x) ]+log [ P (n|x) ], wherein P (r|x) and P (n|x) can be obtained by posterior probability learning of bayesian formulas, i.e., P (r|x) and P (n|x) use a markov chain monte carlo method based on numerical sampling to obtain an approximate solution of the posterior distribution.
The smaller the value of J (X) in the Bayesian formula is, the more the required retrieval image X meets the requirements.
Since P (X) is typically a constant, J (X) can be transformed into:
J(X)'=-log[P(X|R)P(R)]+log[P(X|N)P(N)] (3)
setting a threshold value alpha, and when the obtained J (X)' < alpha, obtaining the detected image which is the insulator image with the dirt.
S5, constructing a Bayes convolutional neural network classifier, and initializing Bayes convolutional neural network parameters by using an Xavier method: a. determining a distribution model of an image characteristic to obtain prior probability and class conditional probability of a sample image; b. calculating the corresponding posterior probability by using a Bayes formula; c. obtaining an image retrieval result set meeting the requirements by utilizing a decision function J (X);
in this embodiment, the specific steps of initializing the bayesian convolutional neural network parameters by using the Xavier method are as follows: a. generating a normal distribution random matrix with the mean value of 0 and the standard deviation of 0.01;
b. judging whether the generated matrix meets the requirement of the formula (1) in the step S4 by using the formula (2) in the step S4, if not, regenerating the matrix, continuing judging until the distribution of the weights meets the consistent distribution of the formula (1), namely
W in the formula (4) is weight distribution, U is uniform distribution, n j The number of neurons in the j-th layer.
x in the formula (5) m For point set x n (n=1, 2,.+ -.) j is a j-th layer neural network, h=0, ±1, ±2,...
S6, training by adopting a Markov chain Monte Carlo method to obtain an insulator image pollution recognition model: training model parameters by adopting a numerical sampling Markov chain Monte Carlo method to obtain probability distribution of neural network parameters, and optimizing the model parameters to obtain an insulator image pollution recognition model;
in this embodiment, the specific algorithm of the markov chain monte carlo method in step S6 is as follows:
consider the sampling process as a markov chain: x is x 1 ,x 2 ,...,x t-1 ,x t ,x t+1 ,. x in the formula t Representing the sample of the t-th sample, the sample of the t+1th sample depends on the sample of the t-th sample and the state transition distribution q (x|x t ). If the plateau distribution of the Markov chain is p (x), then the samples at the plateau of the state follow the p (x) distribution.
Specifically, according to the state transition distribution q (x|x t ) Extracting a sample x 'and estimating a probability A (x', x t ) To take x' as sample x of the (t+1) th sample t+1 。
Due to each time q (x|x t ) Randomly extracting a sample and taking the sample as A (x', x t ) Is accepted, so the modified markov chain state transition probabilityThe method comprises the following steps:
q'(x'|x t )=q(x'|x t )A(x',x t ) (7)
the fine and smooth conditions according to the Markov chain are:
the modified markov chain can reach a plateau and a plateau distribution is p (x).
S7, detecting and identifying pollution types of the sub-images of the insulators: and (3) detecting the verification set in the step (2) by using the insulator image pollution identification model in the step (S6), respectively calculating the probability of pollution and no pollution of the insulator image in the verification set, and selecting the probability with a large value to judge the probability as the pollution type of the insulator image.
The basic principle of step S4 in the present invention is: p (R|X) and P (N|X) are obtained by adopting a Markov chain Monte Carlo method based on numerical sampling, namely, the probability distribution P (X) of parameters is generated by randomly sampling samples from real distribution, and then the probability distribution P (X) is used for searching the expectation of neural network parameter hiding variables.
In addition, in the invention, in the step S4, the Xavier method is adopted to initialize the neural network parameters, so that the convergence speed of the neural network is high, and the model training speed is improved. In order to make the information in the network flow better, the distribution difference between the sample space and the class space, i.e. the probability density difference, cannot be too large, so that the variance of the output of each layer should be equal as much as possible.
In order to verify the effectiveness of the method provided by the invention, the MNIST handwriting digital image sample set and the insulator pollution image sample set disclosed under the python frame are used, and compared with a support vector machine method and a variable decibels method in the aspect of classification effect accuracy, the result is shown in the following table:
from the above table, it can be seen that the classification accuracy of the three methods for MNIST data sets is substantially identical, but the classification accuracy of the markov chain monte carlo method for the insulator image pollution data set is higher than that of the other two methods. Therefore, the Monte Carlo method based on the Markov chain can accurately and efficiently identify and classify the pollution of the sub-images of the insulator.
As shown in fig. 5, the variant db She Sifa and markov chain monte carlo processes are compared: when the model training iteration number epoch is lower than 30, the accuracy of the Markov chain Monte Carlo method is slightly lower than that of the variational Bayesian method; when the epoch is larger than 70, the accuracy of the Markov chain Monte Carlo method tends to be gentle, the model convergence speed is faster than that of the variable decibels Bayesian method, and finally the classification accuracy of the model is slightly higher than that of the Yu Bianfen Bayesian method. Therefore, the Markov chain Monte Carlo method can be used for rapidly and accurately identifying the pollution of the sub-images of the insulator and efficiently judging the pollution types of the sub-images.
The above examples are provided for illustrating the preferred embodiments of the present invention and are not intended to limit the scope of the present invention, and various modifications and improvements made by those skilled in the art to the technical solution of the present invention should fall within the scope of protection defined by the claims of the present invention without departing from the principle and spirit of the present invention.
Claims (4)
1. The method for identifying the insulator image pollution based on the Markov chain Monte Carlo is characterized by comprising the following steps of:
s1, establishing an insulator image library: collecting and storing the data of the insulation sub-images of the distribution line by using unmanned aerial vehicle inspection;
s2, dividing an insulator image in an insulator image library of the distribution line into a training set and a verification set;
s3, establishing an insulator image pollution identification sample set: manually screening the training set in the step S2, and classifying whether the insulator image has a pollution mark or not to obtain an insulator image pollution identification sample set, namely an insulator image pollution sample set and an insulator image pollution-free sample set;
s4, establishing an insulator image pollution identification feature set: each insulator image in the insulator image pollution identification sample set is subjected to segmentation pretreatment, feature extraction and selection, and the features are combined into a feature vector Y to obtain an insulator image pollution identification feature set, and the insulator images are divided into a polluted insulator image and a non-polluted insulator image according to the feature vector Y;
s5, constructing a Bayes convolutional neural network classifier, and initializing Bayes convolutional neural network parameters by using an Xavier method: a. determining a distribution model of an image characteristic to obtain prior probability and class conditional probability of a sample image; b. calculating the corresponding posterior probability by using a Bayes formula; c. obtaining an image retrieval result set meeting the requirements by utilizing a decision function J (X);
s6, training by adopting a Markov chain Monte Carlo method to obtain an insulator image pollution recognition model: training model parameters by adopting a numerical sampling Markov chain Monte Carlo method to obtain probability distribution of neural network parameters, and optimizing the model parameters to obtain an insulator image pollution recognition model;
s7, detecting and identifying pollution types of the sub-images of the insulators: and (3) detecting the verification set in the step (2) by using the insulator image pollution identification model in the step (S6), respectively calculating the probability of pollution and no pollution of the insulator image in the verification set, and selecting the probability with a large value to judge the probability as the pollution type of the insulator image.
2. The method for identifying the filth of the image of the insulator based on the markov chain monte carlo according to claim 1, wherein the specific operations in the step S5 are as follows:
the adopted insulator image pollution sample set is R * ={R 1 ,R 2 ,...,R k The sample set of the insulator image without pollution is N * ={N 1 ,N 2 ,...,N k Probability can be estimated using statistical methods:
the probability of the insulator image being dirty is that the image X is:
P(X|R)≈P(X|R * ) (1)
the probability of the image X being an insulator image without contamination is as follows:
P(X|N)≈P(X|N * ) (2)
according to the Bayesian formula: j (X) = -log [ P (r|x) ]+log [ P (n|x) ], wherein P (r|x) and P (n|x) can be obtained by posterior probability learning of bayesian formulas, i.e., P (r|x) and P (n|x) use a markov chain monte carlo method based on numerical sampling to obtain an approximate solution of the posterior distribution.
The smaller the value of J (X) in the Bayesian formula is, the more the required retrieval image X meets the requirements.
Since P (X) is typically a constant, J (X) can be transformed into:
J(X)'=-log[P(X|R)P(R)]+log]P(X|N)P(N)] (3)
setting a threshold value alpha, and when the obtained J (X)' < alpha, obtaining the detected image which is the insulator image with the dirt.
3. The method for identifying the filth of the image of the insulator based on the markov chain monte carlo according to claim 1 or 2, wherein the specific steps of initializing the bayesian convolutional neural network parameters by using the Xavier method in the step S5 are as follows:
a. generating a normal distribution random matrix with the mean value of 0 and the standard deviation of 0.01;
b. judging whether the generated matrix meets the requirement of the formula (1) or not by using the formula (2), if not, continuously judging the regenerated matrix until the distribution of the weights meets the consistent distribution of the formula (1), namely
W in the formula (4) is weight distribution, U is uniform distribution, n j The number of neurons in the j-th layer.
wherein x is as described above m For point set x n (n=1, 2,.+ -.) j is a j-th layer neural network, h=0, ±1, ±2...
4. The method for identifying the filth of the image of the insulator based on the markov chain monte carlo according to claim 1, wherein the specific algorithm of the markov chain monte carlo method in the step S6 is as follows:
consider the sampling process as a markov chain: x is x 1 ,x 2 ,...,x t-1 ,x t ,x t+1 ,. x in the formula t Representing the sample of the t-th sample, the sample of the t+1th sample depends on the sample of the t-th sample and the state transition distribution q (x|x t ). If the plateau distribution of the Markov chain is p (x), then the samples at the plateau of the state follow the p (x) distribution.
Specifically, according to the state transition distribution q (x|x t ) Extracting a sample x 'and estimating a probability A (x', x t ) To take x' as sample x of the (t+1) th sample t+1 。
Due to each time q (x|x t ) Randomly extracting a sample and taking the sample as A (x', x t ) The probability of accepting, the modified markov chain state transition probability is:
q'(x'|x t )=q(x'|x t )A(x',x t ) (7)
the fine and smooth conditions according to the Markov chain are:
the modified markov chain can reach a plateau and a plateau distribution is p (x).
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111029250.4A CN116029957A (en) | 2021-10-25 | 2021-10-25 | Insulator image pollution identification method based on Markov chain Monte Carlo |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111029250.4A CN116029957A (en) | 2021-10-25 | 2021-10-25 | Insulator image pollution identification method based on Markov chain Monte Carlo |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116029957A true CN116029957A (en) | 2023-04-28 |
Family
ID=86074574
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111029250.4A Pending CN116029957A (en) | 2021-10-25 | 2021-10-25 | Insulator image pollution identification method based on Markov chain Monte Carlo |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116029957A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116416245A (en) * | 2023-06-08 | 2023-07-11 | 国网山东省电力公司电力科学研究院 | Insulator string pollution identification method and system |
-
2021
- 2021-10-25 CN CN202111029250.4A patent/CN116029957A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116416245A (en) * | 2023-06-08 | 2023-07-11 | 国网山东省电力公司电力科学研究院 | Insulator string pollution identification method and system |
CN116416245B (en) * | 2023-06-08 | 2023-08-18 | 国网山东省电力公司电力科学研究院 | Insulator string pollution identification method and system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112435221B (en) | Image anomaly detection method based on generated countermeasure network model | |
CN111353413B (en) | Low-missing-report-rate defect identification method for power transmission equipment | |
CN108229550B (en) | Cloud picture classification method based on multi-granularity cascade forest network | |
CN108562589B (en) | Method for detecting surface defects of magnetic circuit material | |
CN109118479B (en) | Capsule network-based insulator defect identification and positioning device and method | |
CN107341499B (en) | Fabric defect detection and classification method based on unsupervised segmentation and ELM | |
CN109272500B (en) | Fabric classification method based on adaptive convolutional neural network | |
CN112446429B (en) | CGAN (Carrier grade Access network) -based routing inspection image data small sample expansion method | |
CN111832608B (en) | Iron spectrum image multi-abrasive particle identification method based on single-stage detection model yolov3 | |
CN110929848B (en) | Training and tracking method based on multi-challenge perception learning model | |
CN110599459A (en) | Underground pipe network risk assessment cloud system based on deep learning | |
CN112347930B (en) | High-resolution image scene classification method based on self-learning semi-supervised deep neural network | |
CN114359695A (en) | Insulator breakage identification method based on uncertainty estimation | |
CN115471463A (en) | Solar cell defect detection method based on countermeasure type generation network | |
CN116309483A (en) | DDPM-based semi-supervised power transformation equipment characterization defect detection method and system | |
CN113421223B (en) | Industrial product surface defect detection method based on deep learning and Gaussian mixture | |
CN107529647B (en) | Cloud picture cloud amount calculation method based on multilayer unsupervised sparse learning network | |
CN109740607B (en) | Incomplete region detection and incomplete character image restoration method for character image | |
CN116029957A (en) | Insulator image pollution identification method based on Markov chain Monte Carlo | |
CN112990357B (en) | Black box video countermeasure sample generation method based on sparse disturbance | |
CN111223126A (en) | Cross-view-angle trajectory model construction method based on transfer learning | |
CN111797732B (en) | Video motion identification anti-attack method insensitive to sampling | |
CN114694019A (en) | Remote sensing image building migration extraction method based on anomaly detection | |
CN113554685A (en) | Method and device for detecting moving target of remote sensing satellite, electronic equipment and storage medium | |
CN112132835A (en) | SeFa and artificial intelligence-based jelly effect analysis method for photovoltaic track camera |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |